=Paper=
{{Paper
|id=None
|storemode=property
|title=Socio-Semantic Network Data Visualization
|pdfUrl=https://ceur-ws.org/Vol-870/paper_4.pdf
|volume=Vol-870
}}
==Socio-Semantic Network Data Visualization==
https://ceur-ws.org/Vol-870/paper_4.pdf
Socio-semantic network data visualization
Alexey Drutsa1,2 , Konstantin Yavorskiy1
1
Witology
alexey.drutsa@witology.com, konstantin.yavorskiy@witology.com
http://www.witology.com
2
Moscow State University, Dep. of mech. & math.
http://www.math.msu.su
Abstract. The paper is devoted to some information visualization prob-
lems arising in the course of the development of the software package
WitoAnalytics that enables to analyze and visualize data resulting from
the socio-semantic network of the Witology web-platform. The work on
the software is in progress. The article contains a short overview of the
first software capabilities to visualize some types of socio-semantic net-
work subgraphs.
Keywords: information visualization, data visualization, socio-semantic
network, graph
1 Introduction
The Witology company is engaged in solution of some specific real-world prob-
lems by constructing active human community, while developing collective mind
of participants. In order to achieve the goal a collaborative software platform
has been developed and used in the company. Its essential difference from other
similar systems consists in direct involvement of specifically trained facilitators
in the community. In connection with a possible large number of community
members, there is a need of a visual representation of data on their activity on
the platform. The data can be used by facilitators as analytical material allowing
them to quickly make the right decisions.
At the present time, there is a large number of software designed for anal-
ysis and visualization of social networks data (Social Network Analysis, SNA).
They include both wide-section programs to analyze all kinds of graphs such
as, for example, UCINet [3], Pajek [4] and Cytoscape [5], and programs for the
text analysis, for example, Discourse Network Analyzer [6] and AutoMap [7].
Furthermore the class of SNA programs includes specialized software for the
analysis of social networks, for example, NodeXL [8], which allows you to re-
trieve, analyze and visualize data from networks such as Twitter and Facebook.
Since the Witology platform is a socio-semantic network [2], then it requires a
special analysis software package, adjusted to analysis and visualization of this
type of network. Note that the main focus of the research is the scientific field
named information visualization [9, 10], rather than technological problems of
implementation of various methods.
2 Alexey Drutsa & Konstantin Yavorskiy
2 Problem statement
In the paper [2] a general model of socio-semantic network is defined as a triplet
G = (G, C, A), such that
– G = {V, E1 , . . . , Ek ; π, δ1 , . . . , δk } is a social network — weighted oriented
multi-graph, where V is a set of network members, E1 , . . . , Ek ⊂ V × V are
different relations between the members, π : V → Π is a user profile function
and δi : Ei → ∆i (i ∈ {1, . . . , k}) denotes parameters of corresponding
relation;
– C = {T, R1 , . . . , Rm ; θ, γ1 , . . . , γm } is a content multi-graph, where T is a
set of all generated content elements (texts, media, evaluations, tags etc),
R1 , . . . , Rm ⊂ T × T are different relations between the content elements, θ :
T → Θ denotes a function that corresponds to content element parameters
and γi : Ri → Γi (i ∈ {1, . . . , k}) denotes parameters of corresponding
relation;
– A ⊂ V ×T is a authorship relation between the social graph and the content.
For such graph analysis the following task is posed: to develop a series of
visualizations for the most significant activity of the participants in the platform
that would convey the activity in the most informative manner. For instance it
could be user evaluations, text generation and etc. Such visualizations should
demonstrate both time slices of the database and data change over time.
3 Results
In order to solve the task a specialized software package (hereinafter referred to
WitoAnalytics) was developed. As mentioned above, the software developed by
paper authors could be regarded as one of many SNA softwares, but adjusted
to the analysis and visualization of a particular type of graph — socio-semantic
network of the Witology platform. The network presented in the current article
has more than 500 members and, but the visualizations contains around 200
major network members. At the moment the package allows you to build multi-
ple WitoAnalytics monocot graph visualization and visualization of a bipartite
graph.
3.1 User estimation graph
Consider the following oriented weighted subgraph of socio-semantic network:
Ge = {Ve , Ee , δe }, where δe : Ee → [−k, k] × N is a bidimensional edge weight,
the first component corresponding to the average value of vertex estimates (in
some range [−k, k]) and the second component corresponding to the number of
the estimates. Hereinafter the subgraph will be called as user estimation graph.
Such graph could result from user content estimation data taking into account
the author relations for the estimations and for the content which is estimated
(like texts, etc).
Socio-semantic network data visualization 3
Fig. 1. (A) — The visualization ”Elka” (spruce, rus.), (B) — visualization of a local
user neighborhood
Fig. 2. Scaled-up part of the visualization ”Elka”
The two following visualizations of user estimation graph are proposed. The
first is a bipartite representation, where each element from Ve is associated with
two nodes situated on a plain, their vertical coordinates being equal. In this case
the direction of edges coincides with direction of horizontal axis. The visualiza-
tion is named as ”Elka” (spruce, rus.) and its example is presented in figure 1
(A). Here the edge thickness corresponds to the number of estimates between
nodes, and the edge color corresponds to the average value, diagonal edges being
marked out with special color. Histograms of out-estimate distribution (on the
left) and in-estimate distribution (on the right) are displayed near the nodes.
A local user neighborhood of the user estimation graph is presented in figure 1
(B), that is only the edges connected with a fixed user are displayed and the
nodes without visible edges are removed. Figure 2 contains a scaled-up part of
the visualization ”Elka” presented in figure 1 (A).
The second variant of user estimation graph visualization is a monocot repre-
sentation, where each element from Ve is associated with only one node situated
on a circumference. In order to distinguish in-edges and out-edges for a node all
4 Alexey Drutsa & Konstantin Yavorskiy
the in-edges have same joining angle to the node and all the out-edges have an-
other same joining angle, in-angle and out-angle being not coincided and defining
directions, that are symmetric with respect to the radius connected the node.
The visualization is named as ”Solntse” (sun, rus.) and its example is presented
on the figure 3.
Fig. 3. The visualization ”Solntse” (sun, rus.)
Fig. 4. Scaled-up part of the visualization ”Solntse”
The visualization ”Elka” allows us to quickly and accurately provide overall
picture of estimations between users, and to identify the nature of evaluations
of individual users stood out against a background of other users. Thus, for
instance, one can see in figure 1, that all users on average have neutrally esti-
Socio-semantic network data visualization 5
mated each other. At the same time, some nodes stand out among them, their
estimates are almost completely negative, or, conversely, are positive. Such users,
for example, may be taken under special control by facilitators. In addition, such
visualization could be used in order to instantly find a negative evaluation con-
spiracy of a user group against an individual node. This would be expressed in
several broad red lines, leading to one of the nodes in the right column, and
other its in-edges on average would not have red color.
Unfortunately, the visualization ”Elka” cannot identify so-called ”mark up”
groups, in which an agreement between users on mutual positive estimation ex-
ists. Thus, even a group with two members must be a kind of thick green inter-
secting edges in the visualization, their symmetry check is quite time-consuming
process for a large amount of nodes. To solve this problem the visualization
”Solntse” can be very suitable, because in this case incoming and outgoing edge
ends of a node coincide.
3.2 Idea support graph
Let’s consider a restriction of socio-semantic graph Ḡ = (Ḡ, C̄, A), where content
C̄ contains only one relation R̄, which is strict partial order relation on the set
T̄ , and Ḡ contains also only one relation Ē induced by the ratio of A as follows:
v Ēw ⇐⇒ ∃t, τ ∈ T̄ | vAτ ∧ wAt ∧ tR̄τ ∧ τ ∈ T̄ 0 ,
where T̄ 0 — the set of all maximum elements from T̄ relatively R̄. Then such
subgraph Ḡ will be called as idea support graph. Idea support graph is visualized
by WitoAnalytics as follows. The nodes V̄ are allocated on an outer concentric
circumference, and the nodes T̄ 0 are allocated on an inner concentric circumfer-
ence. Size of the nodes and their deviation from the line of the circumference
corresponds to the number of edges. The visualization is named as ”Glaz” (eye,
rus.) and its example is presented in figure 5.
3.3 Short review of current WitoAnalytics capabilities
In the current state WitoAnalytics has the following list of capabilities:
– visualizations of user text estimation (5 types, that include both general view
of the graph, and individual user view);
– visualizations of user actions like ”content creation”, ”content evaluation”,
”content commenting” and etc;
– visualization of user group diversity (dendogram visualizations, adjacency
matrix visualizations, histograms and densities);
– valued graph clusterization (3 methods, that include random max-clique
search algorithm);
– N-gram and word extracting from user content.
6 Alexey Drutsa & Konstantin Yavorskiy
Fig. 5. The visualization ”Glaz” (eye, rus.)
4 Prospect
Since Witology is a relatively young company the work on the analysis and
visualization of socio-semantic network data of the platform is the unfinished
project, in the framework of which one has to solve many analytical problems
and problems of visualization known as information visualization problems [10].
They include the following questions:
– what data to visualize, for example, to detect collusion and ”mark up” groups
of participants for many different subgraphs of the platform;
– how to place nodes and edges;
– which thresholds and for which the parameters of nodes and edges should
be set.
References
1. Alexey Drutsa, Konstantin Yavorskiy, Visualizatsia dannikh sociosemanticheskoy
seti, Dokladi po komputernim naukam i informatsionnim tekhnologiyam, Natsion-
alniy Otkritiy Universitet ”INTUIT”, Moscow, 1, 2012, pp.112-118 (in russian).
2. Rostislav Yavorskiy, Research Challenges of Dynamic Socio-Semantic Networks,
http://www.witology.com.
3. Steve Borgatti, Martin Everett and Lin Freeman, UCINET, Analytic Technologies,
http://www.analytictech.com/ucinet/.
4. Pajek, http://vlado.fmf.uni-lj.si/pub/networks/pajek/.
5. Cytoscape, http://www.cytoscape.org/.
Socio-semantic network data visualization 7
6. Philip Leifeld, Discourse Network Analyzer, http://www.philipleifeld.de/
discourse-network-analyzer/.
7. Auto Map, Casos, http://www.casos.cs.cmu.edu/projects/automap/.
8. NodeXL, CodePlex, http://nodexl.codeplex.com/.
9. Z. V. Apanovich, Metodi visualizatsii informatsii - naukoemkoe napravlenie IT,
Komp’uternie instrumenti v obrazovanii, No 2, 2010 (in russian).
10. Z. V. Apanovich, Ot risovania graphov k visualizatsii informatsii, (preprint)
Novosibirsk, 27 p., 2007 (http://www.iis.nsk.su/files/preprints/148.pdf) (in
russian).